pytorch tensor.norm 挙動がわからない

pytorch の tensor.normをnumpy配列に変換したい

pytorchのtensor.normをnumpy配列に変換したいのですが、tensor.normの挙動がわかりません。
そもそもpytorchのソースコードをchainerに移植したいがため、tensorデータをnumpy配列に変換したいと考えています。
torch.norm(1,3)という処理をnumpy.ndarrayのデータ型の処理ではどう変換するにはどうしたらいいですか？

発生している問題

Exception in main training loop: 'numpy.ndarray' object has no attribute 'norm'
Traceback (most recent call last):
  File "/usr/local/lib/python3.5/dist-packages/chainer/training/trainer.py", line 316, in run
    update()
  File "/usr/local/lib/python3.5/dist-packages/chainer/training/updaters/standard_updater.py", line 175, in update
    self.update_core()
  File "/usr/local/lib/python3.5/dist-packages/chainer/training/updaters/standard_updater.py", line 187, in update_core
    optimizer.update(loss_func, *in_arrays)
  File "/usr/local/lib/python3.5/dist-packages/chainer/optimizer.py", line 800, in update
    loss = lossfun(*args, **kwds)
  File "/usr/local/lib/python3.5/dist-packages/chainer/link.py", line 294, in __call__
    out = forward(*args, **kwargs)
  File "/usr/local/lib/python3.5/dist-packages/chainer/links/model/classifier.py", line 143, in forward
    self.y = self.predictor(*args, **kwargs)
  File "/home/twn/model.py", line 121, in __call__
    x = self.conv1(x)
  File "/usr/local/lib/python3.5/dist-packages/chainer/link.py", line 294, in __call__
    out = forward(*args, **kwargs)
  File "/home/twn/model.py", line 86, in forward
    self.W.data = Ternarize(self.W.data)
  File "/home/twn/model.py", line 32, in Ternarize
    output = np.zeros(tensor.shape)
  File "/home/twn/model.py", line 68, in Delta
    delta = 0.7 * tensor.norm(1,3).sum(2).sum(1).div(n)
Will finalize trainer extensions and updater before reraising the exception.
Traceback (most recent call last):
  File "train_mnist.py", line 133, in <module>
    main()
  File "train_mnist.py", line 125, in main
    trainer.run()
  File "/usr/local/lib/python3.5/dist-packages/chainer/training/trainer.py", line 349, in run
    six.reraise(*exc_info)
  File "/usr/local/lib/python3.5/dist-packages/six.py", line 693, in reraise
    raise value
  File "/usr/local/lib/python3.5/dist-packages/chainer/training/trainer.py", line 316, in run
    update()
  File "/usr/local/lib/python3.5/dist-packages/chainer/training/updaters/standard_updater.py", line 175, in update
    self.update_core()
  File "/usr/local/lib/python3.5/dist-packages/chainer/training/updaters/standard_updater.py", line 187, in update_core
    optimizer.update(loss_func, *in_arrays)
  File "/usr/local/lib/python3.5/dist-packages/chainer/optimizer.py", line 800, in update
    loss = lossfun(*args, **kwds)
  File "/usr/local/lib/python3.5/dist-packages/chainer/link.py", line 294, in __call__
    out = forward(*args, **kwargs)
  File "/usr/local/lib/python3.5/dist-packages/chainer/links/model/classifier.py", line 143, in forward
    self.y = self.predictor(*args, **kwargs)
  File "/home/twn/model.py", line 121, in __call__
    x = self.conv1(x)
  File "/usr/local/lib/python3.5/dist-packages/chainer/link.py", line 294, in __call__
    out = forward(*args, **kwargs)
  File "/home/twn/model.py", line 86, in forward
    self.W.data = Ternarize(self.W.data)
  File "/home/twn/model.py", line 32, in Ternarize
    output = np.zeros(tensor.shape)
  File "/home/twn/model.py", line 68, in Delta
    delta = 0.7 * tensor.norm(1,3).sum(2).sum(1).div(n)
AttributeError: 'numpy.ndarray' object has no attribute 'norm'

該当のソースコード

python3
1def Delta(tensor):
2    n = tensor[0].size
3
4    ### kokokara
5    if(len(tensor.shape) == 4):     #convolution layer
6        delta = 0.7 * tensor.norm(1,3).sum(2).sum(1).div(n)
7    elif(len(tensor.shape) == 2):   #fc layer
8        delta = 0.7 * tensor.norm(1,1).div(n)
9    return delta

試したこと

print(torch.norm(1,3))
tensor([[[0.4926, 0.3223, 0.1846, 0.4800, 0.7103]],

        [[0.7363, 0.3522, 0.7073, 0.2541, 0.4809]],

        [[0.5639, 0.5595, 0.5163, 0.6157, 0.5934]],

        [[0.4394, 0.5042, 0.2208, 0.5893, 0.4093]],

        [[0.5786, 0.4757, 0.6647, 0.4148, 0.5607]],

        [[0.6101, 0.5895, 0.4671, 0.3420, 0.6011]],

        [[0.3711, 0.7060, 0.5985, 0.4870, 0.5184]],

        [[0.6579, 0.6338, 0.5116, 0.5239, 0.5478]],

        [[0.5878, 0.6335, 0.5059, 0.3094, 0.4766]],

        [[0.5310, 0.5487, 0.2004, 0.5366, 0.5692]],

        [[0.7885, 0.4017, 0.6187, 0.3296, 0.1741]],

        [[0.8681, 0.7011, 0.4319, 0.5387, 0.4889]],

        [[0.4878, 0.3948, 0.4182, 0.4235, 0.4727]],

        [[0.3291, 0.5060, 0.6954, 0.4702, 0.7384]],

        [[0.5221, 0.4326, 0.6694, 0.5442, 0.5455]],

        [[0.5701, 0.5102, 0.3249, 0.4733, 0.5535]],

        [[0.5284, 0.5153, 0.4548, 0.5764, 0.4006]],

        [[0.5607, 0.6280, 0.6552, 0.5272, 0.5736]],

        [[0.5310, 0.4631, 0.4940, 0.5698, 0.5756]],

        [[0.6822, 0.4891, 0.5526, 0.5889, 0.3975]],

        [[0.4742, 0.5606, 0.1695, 0.6110, 0.4578]],

        [[0.5681, 0.6335, 0.4142, 0.6283, 0.5202]],

        [[0.3707, 0.5130, 0.3380, 0.5612, 0.6171]],

        [[0.4762, 0.7896, 0.4459, 0.5189, 0.5165]],

        [[0.3796, 0.3661, 0.4132, 0.7394, 0.2711]],

        [[0.4367, 0.5497, 0.2152, 0.2984, 0.4886]],

        [[0.3285, 0.3650, 0.6961, 0.6075, 0.4263]],

        [[0.3891, 0.4031, 0.1322, 0.6820, 0.5218]],

        [[0.4764, 0.5367, 0.2300, 0.4320, 0.5505]],

        [[0.3554, 0.6656, 0.3897, 0.5504, 0.3923]],

        [[0.6184, 0.7238, 0.4366, 0.5892, 0.5281]],

        [[0.4921, 0.4503, 0.4912, 0.6555, 0.4334]]])

(Pdb) p tensor.norm(1,3).shape
torch.Size([32, 1, 5])
(Pdb) p tensor.shape
torch.Size([32, 1, 5, 5])